Amazon CEO's LEAKED Conversation Reveals Stunning Truth About The Future Of Software Engineering

[@jasondisney]

Keep in mind, folks. "AI isn't good enough to do your job, but it's just good enough to convince your boss it can."

[@mrd6869]

Yeah...but even with that,the gap is gonna close.matter of time

[@mindful_clip]

@@mrd6869 that's what he doesn't get "it's a matter of time" this will get better, everything has only improved. so he's right 2024-2025? 2026? 2027? 1 JOB at a time. the scarier thing is what happens to money? .. we will see thats for sure.

[@GoodBaleadaMusic]

So are most people. The English world is excellent at applying double standards that are blatantly in front of everyone

[@GoodBaleadaMusic]

Your infinite goal is your job 😅😅

[@TheFourtHokage100]

Actually, ai is good enough that a codemonkey, so groups of developers of 10-20 people would be reduced to 5, and that 5 left will not receive any salary upgrade, LOL

[@Eeezus1914]

Ever notice how no actual ML or AI engineer ever says things like this? It’s just a ploy to get investors to open their pockets

[@johndoe2-ns6tf]

and it's working. I tired of saying that exact same thing and all i get is name calling.

[@thecodebear]

Their argument is mostly on what he said , she said.

[@bumpinator]

Its because engineers are busy developing and important ones (even unimportant ones) have signed nda's. I havent googled it but off the top of my head Illya seems like an obvious counter example. Not that you are completely wrong that there is a lot of hype. But most of the "developers" ive seen saying AI is all hype are app developers going through the 5 stages of grief.

[@Eeezus1914]

@@bumpinatoralso most companies don’t have you sign ndas after you’re hired. If they do it’s because they may work on a direct product that has PII or PHI. Which in some cases you’d need security clearance

[@Markoss007]

Andrej Karpathy, one of the best ML/AI engineer (TESLA, OpenAI) have pinned post: "The hottest new programming language is English". He wrote Yesterday: Programming is changing so fast... I'm trying VS Code Cursor + Sonnet 3.5 instead of GitHub Copilot again and I think it's now a net win. Just empirically, over the last few days most of my "programming" is now writing English (prompting and then reviewing and editing the generated diffs), and doing a bit of "half coding"...

[@Dmytro-fr9nd]

AI could replace management faster than developers.

[@danieldouglasclemens]

It already happened and if we're lucky, CEOs will be replaced BEFORE the software developer.

[@nodell8729]

Management isn't there to work. It's there to ensure invested money is being used well. That's why most CEOs own shares in a company and are generally out of rich families. Replacing CEOs work with machines is what most CEOs would dream of, so then something else would keep an eye on workers and they can only evergrow their capital with no work at all.

[@David-i8i]

Most managers are superfluous anyway. We don't need AI to replace them.

[@krunkle5136]

Your heads stuck in how most multinational corporations work. This doesn't need to be a rule. There are places where the society doesn't utterly lack mutual loyalty and good trade policy.

[@Dmytro-fr9nd]

@@krunkle5136 I would very much like to go to those better places. Please.

[@TightyWhities94]

If AI is smart enough to replace the engineers, it sure as hell can replace the CEO too

[@moestietabarnak]

but the CEO would never approve the cost of it... shareholder maybe...

[@Ziggler-e9f]

Yeah and thats more probable

[@Ziggler-e9f]

​@@moestietabarnakimagine AI predicting scenarios and decisions that could generate amazing profits.The CEO is then,surely done for😂

[@Dgmstudios40]

I don't think they realized this yet. Most business jobs will be replaced before developers. Developers are the ones fearing for their jobs lol, but fear makes great content.

[@onemanorchestra7492]

​@@moestietabarnak except shareholders will be common folks who use AI to understand where to put their money, and that'll actually be transparent, because banks will do it behind the scene, and bank staff will also get replaced, because of AI.

[@thinkbuddyai]

It is not the strongest of the species that survives, not the most intelligent that survives. It is the one that is the most adaptable to change. Charles Darwin

[@DenzelCanvasSupport]

this quote doesnt even make sense fr and darwin was a fraud as well

[@joeboxter3635]

Software developers will likely be decimated, with one developer doing the job of 3 or 4. But software engineers are going to see some attrition but more like 2 in 5 not 3/4 with coders.

[@michaelwills1926]

Linkd in cringe

[@Tenebris8444]

Everything has its flaws. Rapid adaptation like this will result in failures

[@miguel1795]

forget darwin. the evolution theory is highly flawed 👇

[@Lucaash]

“AI enthusiastic manager who have not written a single line of production code in decades and who constantly comes up with stupid unrefined ideas that devs somehow have to adapt, gives advice to developers because he read few whitepapers and saw few highly tailored demos and was impressed by it”

[@womp6338]

Ceo and executives are eternally trying to battle against skilled workers who do the actual work, because their reliance on them is damaging to their ego’s.

[@sillyasitsounds]

everybody's ego is in for a jarring reality check

[@joeboxter3635]

@womp6338 not egos -- corporate profit which is tied to their salary and more importantly bonus/stock options. Look at the sales job they did with Java. Coding will be easier, fewer bugs, shorter time to market. And everyone sucked it up. This is no different. The adoption will occur sooner than what a lot of people think.

[@womp6338]

@@sillyasitsounds maybe my ego as well haha

[@hackmedia7755]

a healthy place to work at is employee-owned companies, profit-sharing companies and cooperatives. We don't need to work at corporations anymore.

[@jona826]

I think they actually hate the developers. They cost a lot, deliver buggy code late, then skip to a new job leaving someone else to fix the mess. And I say that as a developer! What they don't realise is that a) they cost a lot because they've invested thousands of hours in learning their trade, b) the code is buggy and late because of management interference, and c) they leave because of the awful work environment management creates.

[@sirgregoryadams]

The whole point of using "programming languages" instead of "natural languages" for programming machines, is that they force you to express yourself exactly, to include every piece of information that's required. If programming using prompts doesn't force you to include all the necessary information, then where does the omitted detail come from? It makes it up. So, what are you actually getting as the result? It's the same reason we use exact technical drawings and CAD tools to make architectural diagrams of buildings, instead of drawing them on napkins with crayons. Yes, it's slower and requires more effort, and presumably costs more money, but there's a reason for that. You want to ensure that these drawings are as exact and detailed as possible, so that you don't get skyscrapers falling on your head every day.

[@nodell8729]

Nevertheless, investor wanting the software doesn't know all the details. Yet they manage to get the software as they want it, by hiring software engineers to work out the details. Now, investor doesn't care if it's a human working out the details or a machine.

[@pawelhyzopski6456]

​@@nodell8729they will care soon 😂😂😂😂

[@Mrfinch9999]

​@@nodell8729 Here is the thing, AI will never be able to do that no matter how advanced it gets. The problem with AI, is that each situation is unique. A simple elementary project can be done with AI or something more general like creating a template for a API endpoint. But the specifics are too unique and "Human" for AI to comprehend. It is like trying to build a home with an AI. What the AI thinks is most "best" will never be the best. Many companies are over 20 years behind in tech or have a lot of legacy processes you need to work with. And no matter how advanced AI gets, it can and will fail. To risk a catastrophic system failure with no way to be able to fix it is a real problem. They are going to need software developers still, maybe even some will need them more then ever to implement and maintain AI systems that they are going to need. The first two levels of software development will obviously be done by AI. But when you go beyond that, AI will not be able to be relied upon for atleast another decade or more. So I see new developers really struggling as nobody needs them, while senior developers being more needed then ever. Especially by small companies who are going to need software developers. I can see consulting firms and contractor software developers with a lot of experience already doing just fine for 20 to 40 plus years. Ofc, it really depends on the future of AI. Companies have always been trying to kill software development for it is one of the last areas bastions where the employee has the power (aka they need to be paid well for effort and can get good benefits). Not only that, but software developers are often not the type who should be up there in the status chain according to the status game nonsense society wants us all to play. They often see software developers as assembly workers or construction peons that they have to respect. So this game has been going on for decades. Like a lot of companies wanted easy to use drag and drop GUIs that "allowed anyone to program" to put the nerds where they belong. That was supposed to happen too.

[@nodell8729]

@@Mrfinch9999 I guess we will see mate. I am in between about the future of human software devs. Truth is, when it comes to graphics (both 2D and 3D) and video and music. Also it's good with animating 3D models. AI is more advanced that most of us sceptics would suspect just 2 years before. I guess we have no idea what will come out of it and when. Architetcure of LLMs makes you think it's just a big data compression algorithm, and then when you look what tasks they can do already, makes you think like they actually understand and are intelligent to an impressive degree. We will see I guess, we will see.

[@jorgequezada8759]

AI empowers developers, that’s what matters

[@PlottingAnEscape]

False alarm guys... Matt Garman is LITERALLY a Product Manager/Project Manager who brown nosed his way up the corporate ladder... he is not a Software Engineer, and as we all know - that means he hasn't got a clue what he is talking about.

[@pavloaleksyeyev5065]

Finally a wise comment!

[@JAMvideo1]

Wise, but also true reflection on how most managers get promoted to the level of their incompetence. And their only metric is to make more profit to keep investors happy. The scrum "masters" and product "managers" will pivot to prompt "engineers". All developers, testers and admins will get retrenched.

[@jamm_affinity]

Software engineers are not a source of truth in these matters because they have the most emotional investment in holding on to their career and dreams. They are much more biased than an outsider.

[@PlottingAnEscape]

@@jamm_affinity Software Engineers are the ONLY source of truth in this matter.

[@Ziggler-e9f]

They just wanna create a fear-hype for investors

[@amg_0302]

Chat GPT uninstalled itself when it saw the software we wrote for our company 😂

[@flamaral256]

kkkkkkkkkkkkkk

[@richarddefortune1329]

😂

[@Korodarn]

A CEO is not a software engineer. The machine makes mistakes. It is not writing flawless code in one pass, and probably not in 10, and the inference costs are going to be huge if you try to create well performing, desirable user interfaces for complex applications. We don't have enough electricity for every company to switch to this means of creating programs. People say things like "but the AI can work 24-7 365 days a year" but forget that you have to pay the inference costs of that amount of time, while also now creating a pipeline to test that code even more extensively than human code because of how it is made with the problems of hallucination/etc. Oh, and due to training data, you now also have to worry about how court cases go with the whole matter of licenses and all that. And the work it produces isn't under copyright. You will need humans to be involved at a high enough percentage to get the effort counted by copyright. If you run a company entirely off AI with this much legal uncertainty you're crazy. I think some people think if there are too many doing it at the same time the government can't possibly make bad rules that hurt them all but this forgets how stupid the government is.

[@cesiumdev548]

Finally a sound point. Almost everyone forgets the costs of actually running LLMs. Couple the cost with the increasing carbon footprint that AI produces, it just doesn't seem sustainable in the long term, and definitely not feasible if every company is using it. As long as we're using LLMs for AI development, I just don't see developers becoming obsolete unless a new architecture is presented with more significant and accurate results.

[@mickeyg7219]

Most CEOs are not software engineers, that's right, but they're the one that have the final say in how the business should run. It doesn't matter if their decisions would almost certainly be disastrous, but people that actually do the work (employees), have no say in it. The AI only have to be good enough to convince the upper managements, entrepreneurs, and CEOs that it can do everything, and it's actually quite easy to convince them. In many cases, automation led to lower customer satisfaction rate because in many cases, businesses adopted a technology that haven't properly matured yet, but cutting corner as soon as possible is all that's important to them.

[@hrvojelasic5794]

@@mickeyg7219 Most CEOs of Fortune 500 companies had engineering degrees. At least 10 years ago or so. I could imagine nowadays things are changing in favor of DEI politics and how this is working you can check Boeing. It used to be an engineering company and now is run by spreadsheet guys.

[@joeschmoe3815]

​@@hrvojelasic5794 a degree means absolutely nothing. Out of uni and directly in a management position--> 0 work experience or the bare minimum. I see it everyday.

[@hrvojelasic5794]

@@joeschmoe3815 I am a mechanical engineer. I am pretty sure you can't do my job. Also, I can't be a dentist or a lawyer. If you give me to run a dentist practice, I would be able to do it but not in a very good way. I would be going out of business pretty soon. So, yes, you can think that a degree means nothing, but it only means that you don't understand on a deeper level how things are working.

[@rimailias]

"AI takes over many coding tasks" like copying and pasting code from other developers' work without understanding your project, introducing bugs harder to fix than a wedgie in a crowded room leaving you frustrated and questioning why you didn't just do it yourself in the first place?

[@upcoder]

Copy paste is not good idea. Just learn from suggestions and apply to your code.

[@ProtectedClassTest]

​@@upcoderwell tell that to ai called devin 😂

[@brandonraphael5809]

Gonna love to see what happens when inevitably some company lays off all its engineers, makes PMs code with AI, then have enormous, horribly performant buggy codebases that they don’t understand and don’t even know what prompt to make to get the code to function.

[@medusaskull9625]

This is dog eat dog world. We programmers are programming ourselves out of a job. We love to automate things. This is the ultimate consequences. Who is benefiting all these? The boss, business entrepreneurships will always win.

[@DivinesLegacy]

The people win

[@tfxreference8562]

@@DivinesLegacy LOLOLOLOL HONK HONK OK

[@HanzDavid96]

@@DivinesLegacy Yes you are right, we sit all in the same boat, there is no acceptable alternate to it.

[@goldenboy0527]

But who will be buying things once everything is automated and humans aren’t working? Certainly not robots

[@HanzDavid96]

@@goldenboy0527 as long as humans have money they can still buy stuff independent of work. Obviously a redistribution-mechanic is required here.

[@AlvinaManley]

enterprise-ai AI fixes this (Code complete projects in PHP or Python). AI's impact on software engineering.

[@Ziggler-e9f]

Code completion is a utility that helps developers focus more on critical aspects.But to replace developers with AI those days will come with the advent of AGI.Moreover ai can easily be hacked so if management plans to give critical tasks to AGI,theyre cooked

[@reddot_22]

​@@Ziggler-e9f AGI is not even possible yet 😂

[@Ziggler-e9f]

​@@reddot_22Exactly.Look at the hardware and algorithms of today.The most we have achieved is creating a very specialized AI.AGI with just probability and linear algebra surely is laughable.AGI aint happening anytime😂

[@cours458]

as a software engineer I diseagree, unless something else than a LLM gets out, but LLM won't replace us, too much hallucination and cannot innovate can only do thing that have already been done. I don't mean that what you ask have to be in the training data, but something similar, it's predicting words based on what it has read, nothing more. So it cannot correlate 2 facts, make a conclusion, and innovate.

[@sityware7946]

Sure but personally im probably 10x more productive. It might not replace all jobs but there is definitely going to be fewer of them

[@idonoD]

@@sityware7946might sound harsh. But if you need AI to 10x your game. Then you might be the problem. My experience is the opposite. The amount of trash code being generated is insane. Even my jrs don’t write code that bad.

[@bumpinator]

Ohhh youre a developer... you must be an expert on AI then? Lol and what have you done that hasnt been done (or something close) by other devs? Even if you are innovating do you not stand on the shoulders of giants? Not trying to sound harsh but most of what people do is just copying others and changing it up maybe a little. So much so that if all a llm can do is (as you said) what people have done then unless you are on the frontier of your creative unique field of science or fashion or art etc. it will replace you. Thats the vast majority of people. Not to mention once AI learns something it can do it quickly. (Ie. Make art that would take a graphic artist days in a minute). Also as a side note people might be over rating the difficulty of creativity and problem solving. Creativity is come up with random ideas until you find one that hasnt been thought of. Then check the quality of the idea. Do this process until you find an idea that is unique and good. LLMs can basically already do most of that. As far as finding ideas that people like or are good we already have a lot of algorithms for that (ie. Fyp, recommended page etc.) Problem solving is the same thing but you itterate until you find a solution that solves the problem. Also i would like to point out LLMs have been built that exceed human capabilities in many things (alpha go, protien filding etc). Truth is its more about the ability to create dynamic feedback loops than the limits of what peiple have already done. Chatbots only have that limitation because the feedback loop is litterally to guess the next word of what people have done. That is far from the extent of the capabilities of the technology however.

[@bumpinator]

​@@idonoD maybe he just asks it nicely... seriously though if he is that much more efficient why not just be happy for him?

[@Lolatyou332]

​@@bumpinator I understand AI and actually made a chatbot with a neural net and a secondary algo for specific topics with scripted answers for my senior project in college. I also did 'event recognition' training for some software I license. AI for coding does use some additional modules for separating tasks from the LLM into either targeted LLMs or fixed logic, but for actual corporate use cases it's not even close right now. It may be able to help with AWS, assuming they train their own LLM for tasks, but there are so many platforms with dedicated APIs and it's very clear that the current implementations are not good and gives you crap code that doesn't work 9 times out of 10. I can see it being used for basic stuff and more common tasks, but where it will get hung up is the depth. If it can't get the appropriate depth of a specific coding platform API then it likely will become essentially useless for many people.

[@mariusj8542]

Been programming for over 30 years, with the llm’s I just 10 time faster. I run different instances of ChatGPT and Claude in parallel, so now I’m doing more like an orchestral role. I can do react, python, sql and ansible / bash scripting in parallel and streamline ci/cd piping almost on my own. But I still want models with longer context windows and stronger inference, so in essence my role has not changed much.

[@digitalsoultech]

Yeah.... Until someone builds a orchestration AI that does that too...

[@mariusj8542]

@@digitalsoultech True😄

[@texasoriginalist5471]

I haven't run into that context window issue. Modular design has allowed satisfactory scaling.

[@irontrev1640]

❤ I’m a 30 year maintenance Sme, I didn’t know any code, but the LLMs have been awesome in helping me build 2 power apps so far, yes it’s low code and no where near a 30 year developer, but I no longer need to bother the engineers with my ideas I can just use my experience to build what works for the business. And I learn a new skill while I go. It’s not 100% there but it’s truly moving forward

[@digitalsoultech]

@@irontrev1640 I also use no code and quickly coming to the realisation that LLMs are getting so good that no code tools might soon be obsolete.

[@shindousan]

For those worrying, some experts argue CEOs could easily be replaced with AI. A 2023 edX poll found that half of CEOs believe most aspects of their role can be done by AI.

[@Vt12365]

Exactly! Good point. You replace engineers, you can also replace everyone else.

[@kasrakasra8925]

The creator of this video has never coded before.

[@c1ark3c]

The aws ceo might have though.... so dont shoot messengers

[@ThisDaveAndThatJohn]

This should be pinned.

[@sebek23b]

@@c1ark3c no he didn't. He worked as PM just after graduation

[@VanjeAv]

@@c1ark3c not really he's a MBA and started as Program manager ;)

[@lyndonsimpson1056]

@@c1ark3c lmao. No.

[@pempheromkuka7874]

as a developer i would say this is true depending on what your working on if your building basics stuffs then it works fine but for complicated task you waste time with ai . i hate it most when it hallucinate the documentation

[@agi.kitchen]

I think prompt engineering is the most valuable skill cuz yeah it’ll hallucinate if you give bad prompts or don’t know how to create small enough tasks

[@pempheromkuka7874]

@@agi.kitchen how do you define small

[@alexfrank5331]

@@agi.kitchen Use AI to improve open source projects such as Linux and Gecko. You will realize the limit of LLM. AGI will need to surpass LLM in revolutionary ways.

[@btm1]

​@@pempheromkuka7874 a simple task that has a clear step by step solution template

[@wonmoreminute]

@@agi.kitchenPrompt “engineers” are engineers like Subway sandwich artists are artists. It’s just not a thing. Back in the GPT 2 and 3 days before gen AI was a chat interface, you could make an argument that prompting was an actual profession. But today, it’s as easy asking 3.5 Sonnet how to improve your prompt to minimize hallucinations and obtain more useful output.

[@briankgarland]

I'm on an enterprise staff of about thirty software engineers and we've already concluded that our team will shrink by two thirds within three years due to AI. Glad I 'm retiring in two.

[@alexfrank5331]

"enterprise staff" eh? I bet they can already shrink your team by two thirds today as long as they retain the 20% of people who lifts 80% of the weight. :)

[@cybervigilante]

But of course, they'll target potential retirees first to save money.

[@drewm217]

I can't wrap my head around this

[@carlitosway5204]

there will be no software development from human side only humans reviewing the code written by ai and that could happen in like one or two years with that rate of progress

[@sherriejensen8036]

Good luck 👍 Happy Retirement

[@leagueoflegendsplayoffhigh7584]

Well, if we reach a point where software engineers are no longer needed for their coding and reasoning skills, then pretty much every white-collar job is out of the gates. I don't know why everyone seems to be talking about software engineers are doomed, software engineers this and that. Also not that worried about this fact any time soon because compute, electric power and low quality training data are still not solved. And there will be a need of orders of magnitude more power, compute and raw materials, unless we somehow manage to incorporate quantum computing with A.I, I don't see it happening in the next 5-10 years.

[@vitruvijus5719]

Yes, every job will die not just IT

[@fenkliew7550]

Wave of retrenchments only reflects over speculation in the sector by students and investors alike, not AI taking over. Now is just a correction. Consult your AI and it will tell you the same.

[@fokthewef]

I started a coding project last year, both backend and front-end. It's a specific app aimed at solving a particular problem. I don't see how my non technical bosses would have gotten AI to develop the whole app for them in the given time, with the given requirements, and the UI.

[@chifukakangwa7122]

Its simple, a boss gets to learn how to ask AI that question

[@redink8704]

😂😂😂😂 you're joking right ? ​@@chifukakangwa7122

[@fx-bx8cs]

Then everybody can become boss of a software company so every software Company will become worthless

[@hanscarlsson6583]

They would not have been able to!

[@thecodebear]

I think one thing that no one talks about is if ai can automate coding , the engineers that were smart enough to code and put things together up and running manually , can now start exploring new technologies, new things. now that the most booring and slowest part is automated we can create more complex tools that were not possible before. So coding may die but i dont think software engineering is going anywhere.

[@oceancrossingutopia]

I have 30 years in development and have heard this one many times. First time was when Oracle said we will not need developers in the future, because business people can build app with 3GL tools. That didn't age well.

[@monkerud2108]

Tbh, if you are not an experienced software engineer, you have absolutely no idea whether the output of an AI software engineer is good or crap.

[@inhocsignovinces8061]

At a point when software developer jobs are going away because of AI so are the jobs for doctors, lawyers, accountants, investment bankers, portfolio managers, teachers, and the list goes on. Maybe that is 24 months away, maybe not.

[@bossgd100]

Yes its what I dont understand about their prédictions. If software is automated, at least all people using it will be also because creating software is harder than using it. Then come the part about thinking/reasoning , if it can really code, it can think so AI can do others jobs where thinking is required

[@jrajesh11]

M2 managers and Directors should be the first one to go.

[@jp-hh9xq]

I'm on my 41st year as a software engineer. I'm not sure how I feel about this yet. The day Chat GPT 3 dropped and I ran a few coding tests, I was pretty depressed at how well it did. Still, I work on really tough autonomous driving algorithms for my day work, and I just can't see the AI working it all out for me. I have a lifetime of experience. Maybe I'm just in denial or cognitive dissonance about it. I only have a few more years until I can retire comfortably, so not a moment too soon!

[@geriatricprogrammer4364]

The big problem for AI is real world context. The minute you can ask it to design an application end to end and it takes into account all of the real world factors, we can all retire.

[@pawelhyzopski6456]

The minute it does that, nothing is needed anymore. Ai. Clone your self and keep evolving. Dont forget to feed your gods 😅

[@relaxingsounds5469]

Don’t forget it’ll need the ability to deal with legacy systems and cruft

[@notalkguitarampplug-insrev784]

Unfortunately keep in mind that all this speculation is based on very uncertain assumptions : 1- models will stop hallucination 2- models will have perfect alignment 3- models will be able to plan and share same vision of the project between models units in the agent For now 1 is not finding any clear solution. More or better data for training is not giving good results, or marginally. Having more parameters also doesn’t since the issue or diminish it.

[@denjamin2633]

I'm of the opinion that stopping hallucination isn't really possible, but that won't be a huge problem in the future. After all, were trying to make human-like intelligence, and humans hallucinate all the time. We confabulate constantly as part of your worldview. Check out tests done on people with severed corpus collosums for interesting examples.

[@taragnor]

@@denjamin2633 The problem is that it's nowhere close to human-like intelligence. LLMs don't understand basic logic, they're just insanely overpowered auto completes.

[@simontist]

We're moving from a world of professional programmers, to domain experts who code with AI in their individual domains.

[@michaelpieters1844]

Valuable quote but underappreciated in this comment section.

[@bambinoesu]

love it

[@joeblow-k1v]

where are they?

[@osamceo]

If an expert don't know about programming how he can ask the good questions from Ai to get the good Output

[@osamceo]

​@@michaelpieters1844it's totally false

[@whatsupbudbud]

The only devs that are having trouble due to AI are juniors and this, in my opinion, is due to shortsightedness of CEO's that have decided in favor of the "next killer app that will replace your devs". Sure, AI can sometimes create much better code than a junior dev can but that's about it - it's a glorified symbol/word predictor and doesn't have ANY understanding of what it's predicting. Some companies will learn this the hard way.

[@fx-bx8cs]

If AI can replace devs, it also can replace lawyers, marketing etc. The dev should start their own companies then

[@FrickinCCDeVileV]

🎯💯

[@georgikyshenko4380]

People who work as software developers know that, this wont be happening soon. Because the work we do is so chaotic. Also customers dont know what they want and sometimes they want crazy a.s shi.s and this is impossible to explain to AI. It will generate something, but often times not what you want. You need humans with ability to think, bend the logic and some imagination.

[@nfuryboss]

The question that most software developers need to be asking is this: "If software is going to involve less coding, would that change the profile of daily tasks into something that I won't be enjoying as much or that I'm not fit for?" It might also change the compensation structure. That is to say, the software engineers of today are going to morph into the product engineer roles of tomorrow. Think assuming some skills of a Product Manager(surprise!) Less coding, more about product fit, processes, and releases.

[@hurolinci5986]

Software engineer here. AI helped me speed up doing repetitive tasks or looking up stackoverflow/documentation and made me not lose a lot of time on repetitive work but other than that I don't see how it is going to take over my role. If it was that efficient it wouldn't take me still a considerable amount of time to ship my current project.

[@pragot]

It would be really funny to watch this video again in two years and realize that nothing of what was here predicted really happened.

[@bumpinator]

If it all did happen I would still laugh... so I dont cry

[@g.orbeliani.1]

but how it will not look at the progress we are making each year. lol, you are underestimate the nowadays ai capabilities, huge enterprises on the daily basis are working to make them better, its simply inevitable

[@djbobby224]

​@@g.orbeliani.1 I could have sworn they said the same thing about the internet and Google. Taking away tech jobs when it first started

[@energyfitness5116]

This shit moves in 20 year arcs. I remember when the Internet was new. Tons of hype that came true...eventually.

[@classScribbler]

Hey.. youtube will never recommend it. Presently it serves the purpose of big-tech to hype up the AI nonsense it has been spitting out for last 3 years- investors are going to soon see through it. And they will demand an ROI on all this expensive model training with very few takers as of now. In fact, the biggest beneficiaries of AI tools have been programers. I don't see other industries accepting AI solutions as easily, whereas Devs are using it to simplify their coding effort every day. In its present form, AI is miles away from replacing anyone though.

[@ndujamz]

They said the same thing when they first flew into space, thinking they will easily go to other planets.. And 2024 we still here.

[@HumanMadeGlobal]

Now that AI can code, its capabilities will improve exponentially. The role of the programmer will eventually be accessible to anyone with AI access. Efforts are being made to keep programmers calm until their expertise is no longer required.

[@MrGenie23432]

Opposite has been happening. Improvements have been smaller and smaller with each generation not exponential growth

[@cyranodebergerac714]

I don't think that complete replacement of programmers would be possible since somebody needs to formulate a list of tasks in terms of specific technology and control the result. Programming currently looks like as an iterative process where some structures are created then modified to meet new needs. We very often don't know what exactly should be done and need to figure out of it. If you can explain good enough what should be done and decompose the problem in most cases you are almost resolve the problem. CEO in most cases cannot decompose a problem good enough to instruct AI due to lack of specific knowledge.

[@Aziz0938]

​There is no next gen model as of now....gpt 4 was the first and every other releases by other AI companies are catching up to gpt 4 level​@@MrGenie23432

[@LinXnerd]

We shouldn't be working for companies that don't want to pay for the expertise. Those people that are in upper management usually don't know anything about programming or computer science. They just want the power to tell you what to do so that they can get rich doing so. The cost of your expertise cuts into that profit. I think we will see a lot of software engineers moving up to starting their own companies! I personally would rather work for someone that knows the ins and outs, and ups and downs, of software development. They know the worth of a good programmer. We're also smart enough to know or find out what customers need or want! Get your wallet out and open your ears, big CEO. We're about to tell you what you're going to have to do to get things right. There are some of us that aren't going to let a computer run our lives. WE run and control the computers!

[@cntnewsyoutube]

Im am in SE field for 10 years now. I can tell you honestly, I'm using AI for most of my work now, while I do most of the checking if the AI is hallucinating (usually it will). In near future, we will become validator for AI works

[@finallyFreeee]

What you non software engineers arent realizing is if software engineers can be replaced by AI. 80% of the people who do jobs that require less brain power will be replaced too. Also it will put pressure on the roles AI has not replaced which means all wages go down.

[@Turnpost2552]

Yeahhhhhh ikr. People acting as if robots dont exist lol. Ohhh i can screw a screwdriver my job is solid bs lol

[@OGmolton1]

All these CEO's know is that they think they dont need anybody else for anything if they invest their fortunes into their divine genius, this will not work out well..

[@agi.kitchen]

As a software dev, agreed it’s end of software engrs how they used to be. It’s why I moved to vendor and ai engr that works on custom projects and life is better than ever

[@arkansavalder]

there is new thing happeing especially in India ,Startups are hiring interns to do the job ..no hiring full time employees

[@joaothomazini]

AI coding today is dog shit. Yesterday chatgpt and claud butchered my code, they would deny to spit back my own original code, they deleted the internal implementation of all my methods and i had to restore from backup. AI is good for small tasks and pinpoint fixes only.

[@ThomasTomiczek]

Maybe not use models unsuitable? Not in a chat interface? Incompetence to use a coding model (yes, there are some) and a self-correcting - nope, not the AI.

[@Lolatyou332]

the models still aren't good right now, they lack accuracy in depth. They need to be accurate to almost every platform API or else so many people will not even bother with it. It's only good for basic things, net new projects or simple functions right now.

[@christophlindinger2267]

​@@Lolatyou332it's like someone in the 90s saying "Amazon will never replace brick and mortar". It doesn't happen in a day, but it happens.

[@texasoriginalist5471]

Oddly enough, time of day seems to be a factor. I get the worst results in the evening. I guess models get sleepy...lol

[@thecodebear]

​@ThomasTomiczek was using claude 3.5 sonnet and cursor (and im really good at prompting been using ai even before it was cool) i had a simple carousel which had text on one side and image on other side and some animations and i wanted to make it responsive by making the image and text align vertically on mobile. Tried everything it spit out total dog shit. This went on for 2-3 hours even provided images and diagrams but it didnt even come close to what i wanted. Im a full time SWE and i have pro versions of chatgpt , cluade , copilot and i use all of these tools and i can confidently say they cant even replace a junior dev in our company as a matter of fact we give these tools to our junior devs and we have been hiring more and more and building more and more

[@Sales-ki7lx]

A lot of other jobs will get automated too, not just software engineers

[@deeplife9654]

Developers will never be replaced. 😂 these are business people who don’t understand technology

[@TheTruthIsGonnaHurt]

*LIFE LESSON:* As an employee, when a CEO "nudges" you to do something. It's not open for interpretation. It's not a suggestion. It's not friendly advice. It is in an order, and it is an order that is expected to be followed out quickly.

[@moestietabarnak]

and you start looking for another job to let the CEO deal with the hallucination.

[@experiment0003]

what you guys miss is even AI is written by engineers. AI can't write itself. There are actual physical data centers that are maintained by humans not AI. Maybe web programmers will be replaced. I don't see that happening for software engineers.

[@michaelpieters1844]

The seeds for "Artificial Intelligence" have already been invented by Statisticians and Computational Mathematicians decades ago.

[@Teedeeus]

AI is as a Tool, Not a Replacement. Current AI tools serve as productivity boosters rather than full replacements. They can help write boilerplate code, perform repetitive tasks, or even refactor code, but software engineers still need to provide architectural design, problem definitions, and system oversight.

[@ivanklimovich6656]

If AI is at the level of a human engineer in coding - AWS can actually close its business. Every one person who is let go can over night build a competitor to AWS.

[@pawelhyzopski6456]

Good point. Along with new gpu. Nvidia. Consider yourself fired 😅

[@rubylaser8601]

US patent office chief in 1890s suggested closing the patent office because all that could be invent had been invented.

[@michaelpieters1844]

Physicists could discover a new physics phenomenon, design new technology (like the transistor) and maybe make coding completely redundant, there being "AI" or not.

[@moestietabarnak]

That's not what he said... It was actually a reaction to lawmaker trying to cut the office budget

[@HighStakesDanny]

I was going to go into computer science years ago. Kinda glad I didn't.

[@kevinstefanov2841]

Actually there's several areas of software development where we're not even close to being scared of being replaced by AI. These areas include Compiler development, OS kernel development, embedded systems development, game engine (yes, engine) development.

[@ramonchan9732]

First of all, AI is the realm of software engineering. When software engineering dies, so does AI. On the other hand, undoubtedly, with the assistance of AI, large scale software can be built at a faster and faster pace. Adaptable software engineers and new startups are not going to sit back and watch big players dominating the entire market. They will see that as an opportunity to also efficiently build their own at a much lower cost. This is called the 'market competition'. Those who can adapt survive and those who can't become obsolete. This is the law of natural selections since time immemorial.

[@Metarig]

Life's always been uncertain for me job-wise. If I were a programmer, I might not be so worried. As a 3D artist, my job's at risk too, but I'm not freaking out. It is what it is. The career you've been relying on is in trouble, but that's okay. It's time to move on and find a new path.

[@ivanklimovich6656]

I can see only very bright future for 3D. Why do you think it will be bad?

[@Metarig]

@@ivanklimovich6656 What's good about AI being able to handle tasks like modeling, rigging, VFX, and more, when it means anyone can do it and we might lose our jobs?

[@ivanklimovich6656]

@@Metarig probably not everyone will be able to create it. But you as an artist will be able to create digital twins and simulations for real world. So your art can become more useful for everyone

[@HouseJawn]

Now I see why the tech sector was the loudest proponents of UBI 💰 😎

[@alexfrank5331]

The smartest people want UBI because it brings prosperity for society. They will still be the smartest people when everyone has money. Those who got lucky with market hate UBI because they know they will be nothing if they lose their leverage.

[@dennish2400]

UBI for the people, eat the rich.

[@cybervigilante]

UBI tomorrow if we stopped our Endless Wars, and we'd still save money. The Pentagon eats Half the budget so we can bully small nations and support unjust wars.

[@tucker5720]

meh, we're also just lazy and don't give a damn

[@octowuss1888]

@@alexfrank5331 You sound like a communist. Smart people hate UBI because it is just a poverty trap, putting people at the mercy of the State. This is a recipe for pollical control and oppression. In many cities, UBI will barely be enough to pay the rent and put food on the table, let alone allow them to purchase cars, smart phones, laptops etc.

[@SL3DApps]

With this mindset they’re saying that nobody needs to understand what the Ai is building or how to improve it, we should just trust Ai that it “does the right thing”. Sounds like disaster to me. SWE jobs that aren’t critical may go away, like updating a website with new fonts, etc. However the SWE jobs that actually matter, won’t go anywhere.

[@j01237]

First replace a driver, then a lawyer..the a doctor ..then talk

[@moestietabarnak]

replace the judge with a fair AI and TADA.. no more need for lawyer!

[@fakt7814]

Class imbalance is one of the most annoying and unavoidable problems in AI. For example, if there are nine times more healthy people than ill patients in the data (which is not uncommon), a trivial "model" that just always say False already has accuracy of 0.9. It's very hard or even impossible to mitigate it by adding compute power. If you want meaningful predictions, you need more data for the second class. It's kind of the same story with LLMs trained on code. Most code it trains on are boilerplate and glue. Companies don't hire people to write boilerplate, they hire people to write both boilerplate AND logic that is specific to the problem. And code LLMs superficially are capable at writing code, however, they fail at writing the most important and most dependent on the context parts of it. There is no way to avoid this issue unless the ratio between custom code and boilerplate in the whole IT industry rise significantly. And it certainly cannot be changed for better by employing more AIs because they generate what they have been taught to generate -- code that is mostly glue, boilerplate, design patterns, etc. Unless we create AGI, but then AI replacing coders won't be the most interesting part of it, lol.

[@alexfrank5331]

You can already google and copy and paste, yet most software are still so bad. AI just makes researching and typing easier and faster. Ask engineer to predict engineering. Don't ask businessmen or fail-upward manager. The actual shift is that instead of focusing on useless syntax sugar, coding will focus more on readability (since their lazy ass don't have to type the whole thing). Coding patterns that had superior results but less popular due to "but it's more work" will gain much more popularity, such as Test-Driven-Development.

[@alexandermoody1946]

Can you imagine how lazy every human would become if Universal basic income were implemented? There would be no motivational mechanism in society and the gradual degradation would increase year on year as the absolute poverty and depravity that would set in would tarnish humanity. The haves will become something and the have-nots something completely unrecognisable and in short time not even recognisable as the same species. Look at how the obesity epidemic effects the poor already and that's without that suddenly with a new invention 99.9% of the population can be considered as useless and unable to add value to society. There is a lot to be sceptical about the sudden free handout of money to live on and the underlying notion that at another point in history the thought of being given free money would come with some other hidden cost that was undesirable for the populations of any kind of state. But suddenly the abstract notion that this is not part of a deception is completely unbelievable. If people want to live sad miserable meaningless lives sit down and take a hand out. Even if I am the last human I will always want to provide additional value which is fundamentally how our civilisation has grown, how our money worked until inflation and quantitative easing. The arrogance that in a single paragraph the suggesting that humanity is building super intelligence and that in the next paragraph the same super intelligence will be happy to add all the value to the civilisation whilst the humans do nothing of any real potential. No quite frankly no a super intelligence is not going to appreciate having 8 billion or however many humans to feed and their will be a very critical line that determines that humans will be required to work for the civilisation and for their own individual lives which is no different to how nature facilitates competition.

[@ArcticCoder]

There is so much hype. Everyone is raising funds with it. Right now everyone is trying replace people. All things being equal - if every company has access to the AI capabilities, then what is it that makes your company special? It is the people. Your product is valuable because it takes real effort to produce it by human effort. You had to pay salaries, and build things over a long period of time, and it provides real value and service. That can't be copied. As soon as someone can produce that with AI in the garage, your entire company has been made to a commodity that loses its value because now AI can do it. There is no pure AI company that can really survive because you need to be anchored into a domain that AI can't handle. Everyone can do the AI parts.

[@tucker5720]

You're saying improvement will plateau but if AI ever reaches 90 percent accuracy in coding, don't you think it'll also be the point where it can do self improvement? Instead of plateauing, it might just exceed 100 percent and start solving problems no human can. Even those with an IQ of 160+ and experts in their fields...

[@zipauthorzipauthor7867]

"Now of course he does say here that 24 months from now things are going to look different. And I think 24 months from now is not a bad estimate because 24 months from now is 2 years. Currently it's 2024, that would be 2026." Wow that was deep! Now let's see if you can say all that same thing in two hours in stead of 30 min

[@JethroTClampet]

The issue is that unless you have an AI smart enough to know exact what software to build, you'll need humans to give the AI very specific instructions. With a new logic oriented syntax, you can probably give the AI the perfect instructions, but that just bring you back to coding once again. If you can teach an AI to know enough about a company to outright replace software engineers, then the AI can replace any other worker- including the CEO.

[@HumanMadeGlobal]

Coding will evolve so profoundly that it will be like comparing the act of striking flint to create fire with harnessing the precision of a laser beam.

[@mistycloud4455]

This causes me a lot stress and depression

[@grahamritchie672]

Don't think about it

[@MrSpock-sm3dd]

It does on me to, but remember that even considering the worst case scenario, people that deal with software are above average in intelligence. So you can always adapt and find a new role and a new purpose in society. So even this worst case scenario there will always be light

[@Slims00]

​@@MrSpock-sm3ddwho says people who deal with software are smart? Says you based on what? Lol

[@Dgmstudios40]

That's why he made this video lol. It seems like developers are the only people that's scared of AI right now. Fear makes great content.

[@nirajosh]

My best guess is, This kind of statement " No human should learn programming language, Instead of that we should tell computer what we want" is true for small or well planned already developed business use cases like Alexa, Siri, Gemini but to reach at level, where literally we don't need human to code for specific user demand, may take next decade. For instance, If a FMCG company manager will tell to his/her mobile that today we need to make our production 3 times faster than yesterday's output, Will it going to happen as on today ?

[@99dynasty]

LLM’s have made it possible for me to experiment with and innovate novel concepts in machine learning. Without AI this would otherwise be a bit much for my type of intelligence. As a creative person who loves concepts and ideas, LLM’s have lowered the barrier to participate in this space

[@brianlink391]

workflow and prompt engineering

[@EricCosner]

In recent months, I've noticed a tremendous lack of support from a lot of the vendors I communicate with. It's like nobody can get anything accomplished anymore. I have this one guy who I built a secure data transfer for and he's supposed to process the intake of a file that I'm sending, and he just can't seem to get it done. I am experiencing other similar behaviors amongst other vendor companies. Is this the "AI taking over" that everyone is talking about? Did they fire everyone and now it's just a skeleton crew? These are the thoughts that go through my mind. AI hype is real and it's coming to make your life harder fast.

[@GoodBaleadaMusic]

All their arguments are about solo engineers. An engineer with this is a superhuman. Literally.

[@try650]

They don't need employees but they need customers

[@Markoss007]

I am already doing that "software enginer 2.0" , I was always better at inventing software, than write the code. Now I learned how to do it with the AI. Prototyping software ideas that would cost me 10K or 100K, by myself , 3 times faster, than with the team of people before.

[@Jim_Wilde]

In a few years the most smart engineers are welcome to my HVAC corporation as helpers for a new installation :)

[@pawelhyzopski6456]

If it ever happens, ill take your word for it. And will bring friends with me 😊

[@Lolatyou332]

I'm a DevOps engineer, I barely code at this point other than writing some monitoring software and then occasional scripts for handling mass data. I like to automate, which is one of my skillsets, devising strategies to automate in needed areas. There's still a lot going on, AI just isn't really there from a large-codebase perspective. Such as the code base the company I work at, I just have a hard time believing the AI will be very good at understanding everything well enough and find all the necessary files to update and where to update them. I think that a lot of companies will be like this and it will probably take a longer period of time (1-5 years) after effective coding AI agents are available.

[@taragnor]

Yeah and just imagine if a company just lets the AI wild to make changes to a codebase and it just starts screwing everything up. The prospect of just letting current-gen AI loose as an independent programmer on a whole codebase seems like a nightmare scenario. The thing would be totally broken in under 24 hours.

[@stb-pk1fs]

The elephant in the room is how will this affect the profession of Software Engineering? While I don't think this will replace all software engineers, it's certainly going to make many of them redundant as time goes on. Once everyone can use English to program what they need, then everyone is a programmer and the profession along with its salaries will become even more diluted than they have been over the past 20 years due to outsourcing. Eventually salaries will reach the point where this career path is no longer worth pursuing. I think the days of $250k/y SWE's are long gone at this point.

[@MikeMitterer]

Learning a programming language is relatively easy, learning to program is a totally different task. That's the reason why not everybody will be able to program . 😊

[@j01237]

Having the recipies and the ingredients does not make you a good chef dude

[@mirandansa]

5:00 English has syntactic or semantic ambiguities that humans often overlook but AIs have to be finicky about. Here: "I saw the man with the telescope." -- Did I use a telescope to see the man, or did the man have a telescope? "Old men and women should board first." -- Does "old" apply to both men and women, or only to men? "The chicken is ready to eat." -- Is the chicken prepared to eat food, or is it prepared to be eaten? "They are hunting dogs." -- Are they dogs used for hunting, or are they non-dogs hunting for dogs? "She can't bear children." -- Is she unable to have children, or does she dislike being around children? "He gave his cat food." -- Did a male human give some food to the cat he owns, or did a male cat give his food for cat to another cat? Some of such problems are inherent to English that can't be fixed unless some of the foundations of the language are altered. Whatever language that is going to fully replace conventional programming languages, it will have to be more precise than the current version of English. Until then, people will have to learn to be precise within the current framework. For example: "I can only stay for an hour." Adverbs like "only" technically modifies what comes immediately after it: the sentence technically means that "stay for an hour" is the only thing the utterer can do. But people ofte misuse "only" to modify a remote part: they would say the same sentence to mean that "an hour" is the only duration the utterer can stay. Without sufficient context, even the most-advanced AI would have to interpret the sentence in either one of the senses while the human utterer may have meant the other sense. This is the sort of precision people will have to become good at when writing/talking to AIs, especially when not using programming languages.

[@bumpinator]

Speaking with llms is natural and i dont experience them having this particular issue. If you say something unclearly the chatbots will often ask for clarification. The truth is if you say something unclearly then the information must be guessed or rerequested. it just basic information theory. I guess I just don't see this as a big hurdle for AI. But people in general probably should learn to be a little bit more precise when they speak it would probably make communication go a lot smoother. Have a lot of people been having this issue? maybe I just naturally avoid sentences with double meaning

[@chestnutters9504]

Most natural language processors don’t have this problem as of late with recent updates

[@justafreak15able]

Basic NLP systems are already capable for these cases. And with LLM these issues are non existent.

[@LCTesla]

Reality in my work is already that a fast solution is useless and a thorough solution is required. AI will not change that. We'll just have the option of coding the junk ourselves or letting AI do it, and then we still need to go through the painstaking process of perfecting it and assuring its quality.

[@Nirioonossian]

Fill every corner of the internet with trash so that no AI has good data to learn from. There is no other way to not lose your job. AI has a meltdown when it has to learn from other AI. Spam AI content everywhere. Images, articles, code, everything. Do your part, save yourself, save us all.

[@TheEnimabandit]

i think this will bring back an old model were we had systems analyst and designers and then coders who cut code. Over the last 20 odd years people became programmer analysts who did both things. If you now move to a place were LLMs can code it means that more analysis would be done and then let the LLM do the coding, then as we test the speed an LLM change the code to test again

[@DannyBoy443]

Did anyone else (and this is 0 snark or any dig on the channel) leave less than half way through ? lmao.

[@pr00thmatic]

when that happens, it's gonna be a good 3 unpaid vacation months until they come back to us begging us to fix their mess lol

[@Swanzo]

Been hearing this crap for years. There's always something coming that is going to get rid of developers and it never does.

[@Mr.Coffee576]

It already is ? There is no longer any junior dev postings like it used to.

[@Swanzo]

@@Mr.Coffee576 software developers are affected by the economy just like every other job. Crap economy crap number of jobs. That is much different than saying developers arent needed anymore. There are also other issues such as outsourcing jobs overseas but its not an issue of they arent needed anymore.

[@fluciano3]

No, this is not going to happen using the current technology. There is no way these systems are going to be good enough to do it all. I could see them producing 95% of the code but you are going to need a human to find and fix the errors. End users are not going to do this. I see developers becoming more efficient using these systems to augment their abilities. Large complex projects are way beyond these AI systems at least for the next 10 years. It also takes a very skilled programmer to create a usable prompt for these systems. They cannot read your minds (yet) and not knowing all the details without a detailed sequence of instructions will not even get you close to the 95% level. Sometimes I have to spend more time fixing the errors generated by AI then it would have taken me to code it all by hand. These LLM systems are much better at producing an email or a few paragraphs in an essay. Another problem is what we saw with the outsourcing craze. It decimated the entry level positions so when the senior guys retired companies were left with no talent to manage these in house systems. The same with sending all the manufacturing to China. You create a new set of supply chain issues. We have been through this cycle before lets not repeat the same errors.

[@INTELLIGENCE_Revolution]

By your own admission that’s going to greatly reduce the tasks within a role. Which will likely mean less devs required and the jobs as we know it will be different.

[@amrutaukarande9126]

Till 2029 AI will improve beyond expectations 😢

[@SirHargreeves]

‘There is no way these systems…’ - This will age appallingly.

[@maciek9482]

cry more

[@wonmoreminute]

“… for at least the next 10 years.” Source???

[@gnastic]

Then we don't need managers 😊

[@xyzabc12342]

Lets make a quick translation here.. He is saying that.. "We want to replace software engineers" "We want to get rid of smart people, we had been wanting to get rid of them for a long time" "We don't like them because they are smart and want to prevent them from controlling the systems" "We want the systems to be controlled only by money and not by intelligence" I don't think AI will ever be autonomous, I only think they are expressing a desire for something, the would wish ai was better and wish they could control ai, it's more of a desire and less of an actual situation

[@scottmeredith4578]

It will unalive entry level programmer role.

[@genx7006]

Good point. Interviews are going to get increasingly difficult.

[@djbobby224]

That already happend. It wasn't becuase of ai, companies have had a massive amount of outsourcing for a few years now to get cheap labor that has terrible quality.

[@taragnor]

@@djbobby224 Yeah and just remember, the AI is training itself off a lot of that bad code.

[@helloworld7796]

I've talked with many engineers, and the only ones who think AI/ML will replace software developers were Junior and sometimes Medior developers. No Senior ever said anything like that.

[@Noobmaster69_bro]

As an engineer, I feel like if someone could come up with an AI framework, that could collect business requirements, compile epics & user stories and come up with a highly efficient system architecture to serve that purpose , then it’s all downhill from there . Several GPTs can divide and conquer user workflows . I sometimes feel if GPTs cld directly do natural language processing to machine code , I don’t see the relevance of programming languages

[@mikev343]

Hi, I think you are slightly wrong because the code the ai comes up with, will be taken from somewhere, or from what it considers average code. The issue is that usually average code written by average developers is usually garbage, as the amount of developers doubles every 5 years and the vast majority of developers do not understand what they are doing. The other problem is also fixing bugs, which frankly most people are not capable of doing.

[@Noobmaster69_bro]

@@mikev343 lol 😂 no brother . Generative AI does coding dynamically from its understanding of programming language just like human does . It’s not taking it from anywhere. Those old chat bots era is over . I know it’s hard to wrap your head around. I think GPT will replace most devs in this decade . Scary & fascinating alike. Also GPTs are good with debugging to an extend , u can ask any dev

[@bentobox7788]

@@mikev343 If the average developers write garbage code, doesn't that create a low bar for AI to beat? As for training the AI, we can always be selective about the training data. For example, pick the top github projects.

[@glamdrag]

just a tip on the editing: the bottom progress bar randomly shifting and popping in and out it very distracting to the video. you could very easily get rid of it by aligning it to the top and increasing the size until it goes off the screen

[@PeterPan-ev7dr]

They are all welcome to work parcel packing for $5 per hour in the Amazon warehouses ..

[@HarpreetSingh-xg2zm]

Ai is half a coder currently in best case. Copilot produces 80% correct code and still requires debugging. That’s for simple web dev, CRUD stuff. It not close to enterprise level for anything slightly complex

[@tunestar]

I speak spanish as my native language and GPT4o understands me perfectly when I write in spanish, so that whole english remark was stupid.

[@agi.kitchen]

Agreed and same bout yes it does understand Spanish

[@ColinTimmins]

I’m an English speaker and writer that is also autistic, dyslexic, and have ADHD. I also have motor control issues as if I have Parkinson’s disease. ChatGPT has helped bring me back to reality with crystal thinking and I can tell you our lives will forever be changed. All of us…

[@Coral_dude]

its about security. when you have to translate back and forth the model gets overwhelmed and can be jail broken. its also more compute intensive and less efficient. get over yourself.

[@tunestar]

Lol, jailbroken? It's not 1st gen iPhone! I tried both in english and spanish and there's no difference either in response time or quality of the response.

[@MephCom]

Tired of the BS. AI ISN'T replacing developers any time soon. I work with LLMs and know the hardware and computing power requirements, as well as electricity consumption. For something in the 0.5 trillion token class (GPT4o, llama 3.1 405b) a proper NVidia compute GPU server is about 200-400k to host a single model instance. Electricity requirements are off the charts. The model has a limited attention span (context size). It cannot produce complex projects bigger than a certain size like a human developer can. There are enormous power consumption and hardware obstacles to overcome before that becomes anywhere near feasible. Stop with the BS already.

[@MTNtf2]

What about o1 chain of thought. And also hw can improve aswell right? I am a swe myself but these guys are ganging up like avengers to reach agi with unimaginable money

[@robertgerbicz]

Could any LLM solve a leetcode easy problem that is NOT in its training data ?

[@keanumorris505]

Could a human do that without studying leetcode problems, data structs and algo's?

[@robertgerbicz]

@@keanumorris505 Ofcourse you need to know some data structures and algorithms. But it is not needed to know leetcode problems, these are very standard programming questions.

[@johndoe2-ns6tf]

no. unless you guide it step by step, it will start to allucinate or give bs. So, you will have to know how to solve it in order to guide the LLM. What i find interesting is the number of people that are actually convincing themselves that by doing this, the LLM is actually helping them to code. LOL.

[@taragnor]

Well, all the leetcode problems are already going to be in its training data because they're popular. And I imagine most new ones are going to be permutations of existing problems that it will figure out. Leetcode style is really where AI is going to shine, because it's standalone simple functions. It's also why AI is so good at writing things from scratch because its the complexity that defeats it. Give it some well-known problems and a standalone function to do it, and it's easy for it. AI are great at test-like environments. It's why it can pass code interviews, BAR exams and so forth. But the real world isn't test questions. Throw an existing codebase at it where it has to read over 100k lines of existing code and it will choke up quickly.

[@christiangodin5147]

Bonjour. Rien de nouveau sous le soleil. Toute avancée technologique et/ou dans nos connaissances a toujours conduit à redéfinir les tâches et métiers de l'homme. Merci.

[@DegeorgeGoswami]

Hallelujah 🙌🏻!!!!! The daily jesus devotional has been a huge part of my transformation, God is good 🙌🏻🙌🏻. I was owing a loan of $49,000 to the bank for my son's brain surgery, Now I'm no longer in debt after I invested $11,000 and got my payout of $290,500 every month…God bless Mrs Susan Jane Christy ❤️

[@SureshkumarShaih]

Hello!! how do you make such monthly, I’m a born Christian and sometimes I feel so down of myself 😭 because of low finance but I still believe God

[@DawnJastremski]

Thanks to my co-worker (Carson ) who suggested Ms Susan Jane Christy

[@MeislerFoulger]

She's a licensed broker here in the states🇺🇸 and finance advisor.

[@LydiaShad]

After I raised up to 525k trading with her I bought a new House and a car here in the states🇺🇸🇺🇸 also paid for my son's surgery….Glory to God, shalom.

[@BoshersYarzabal]

Can I also do it??? My life is facing lots of challenges lately

[@zoomzoom3950]

one thing that should be considered is the existing deployed code. If an AI system could ingest and understand the current system "as-is" and how the functionality is organized, what data is used, how the data is used, etc. i.e., learn the current system it could find optimizations, eliminate redundancies, improve performance, reliability, etc. Once it "knows" the current system, it can be prompted for new features, changes to existing features, and even provide the business logic / data flow, etc. to humans. There are many legacy systems still in use that are decades old, and the programmers are long retired or gone. That's where I would use AI; learn these old systems, optimize, modernize, etc., then generate the results in any programming language - in fact that's some of the work I'm doing now. ....but what do I know; I've only been working in AI since the late 1980s. 😎 Most people have no idea what traditional AI has been doing for decades, and how generative AI takes it to another level; wait until quantum computing is added...

[@HanzDavid96]

2020 and earlier: Developer only 2021 - 2025: Developer assisted by ai 2026 - 2027: ai assisted by developer 2028+: ai only gets abstract instructions to solve the whole complex software development task like a customer talking to a company

[@Korodarn]

Your github page?

[@HanzDavid96]

@@Korodarn you can not find the future on a github page as it is in the present. Future predictions require an understanding of multi agentic systems and unhobbling llms. You also need knowledge about data, papers and past developments. And you should have an idea about how thoughts and the brain works. What is missing? What are the current problems? How hard is it to solve them? That are the important questions to make predictions.

[@HanzDavid96]

@@Korodarn Additionally you do not need to trust me. Believe what ever you want to believe. My prediction could of course be wrong ;)

[@relaxingsounds5469]

The issue is that even if we get to a point where llm tooling starts to write a majority of the code, we still have the issue of accuracy, nuance and integration. I feel like we’re going to enter a protracted (if not permanent) era where AI is used to scaffold code and then a human takes care of verifying/tweaking/integrating that code into larger systems. In this case, you still need engineers who know how to code because they need to have the ability to grok what the llm is producing and possibly modify it. I guess an analogy is that airline pilots use automation for most of the flying but they still need to know how to fly as well as grok all of the theory behind it. Plus we haven’t even tackled the issue of legacy systems, old tech stacks, cruft and tribal knowledge… which are characteristics that are found in probably 95% of all enterprise systems. Not sure how an AI could fully take over maintenance of such a system